Technical note: Comparing ozone production efficiency (OPE) of chemical mechanisms using chemical process analysis (CPA)
Abstract. Chemical mechanisms are critical to chemical transport models for air quality research and policy analysis. Several mechanisms are available and intercomparison, especially using metrics which reduce sensitivity to modeling scenario, is important for interpreting results and assessing uncertainties. Here, we investigate Ozone Production Efficiency (OPE) as a comparison metric under conditions where nitrogen oxides (NOX) are limited. OPE is the net number of ozone molecules produced per NOX molecule lost and can be computed in simulations using chemical process analysis (CPA). We compute OPE (OPE-CPA) for four chemical mechanisms (CB6r5, CB7r1, SAPRC07, RACM2) and find a similar response to varying anthropogenic emissions of volatile organic compounds (VOC) and NOX. RACM2 consistently produces the largest OPE-CPA and differences between mechanisms are greatest at high VOC/NOX ratios. The high RACM2 OPE-CPA is partially due to a slower OH + NO2 rate and potentially to its treatment of NOX recycling. OPE-CPA is generally consistent with aircraft OPE measurements downwind of Houston but direct comparison is difficult due to uncertainties in deposition and VOC speciation. More recent OPE measurements are required to determine whether trends over time are consistent. OPE-CPA responds nonlinearly to NOX and increases at low NOX even as ozone production decreases. Using OPE to predict ozone response to NOX emissions reductions is therefore an oversimplification that will tend to overstate ozone reductions. OPE-CPA is a viable metric to compare mechanisms, however, additional work would be helpful to define standardized conditions for comparisons.
Review of "Technical note: Comparing ozone production efficiency (OPE) of
chemical mechanisms using chemical process analysis (CPA)" Â Tuite et al.
The manuscript presents a useful review of OPE from several field studies in comparison to model OPE results using several widely used photochemical mechanisms as part of a diagnostic analysis of modeled ozone concentration. Â EPA guidance for air quality model evaluations emphasizes that model performance evaluations (MPE) that are based on modeled and observed species concentrations (state variables) are inadequate, and that other types of MPE, including diagnostic evaluation, should be used in addition to the state variable MPE. The diagnostic analysis presented in the manuscript is useful for gaining insight into feedback processes in photochemical mechanisms and air quality models. Diagnostic model evaluations are rarely performed in regulatory applications of air quality models, and this manuscript illustrates how the chemical process analysis (CPA) in CAMx can be used for diagnostic evaluations.
Additionally, as the authors point out, most O3 nonattainment areas in the U.S. have become primarily NOx-limited for peak O3 concentrations, and thus there is a need for more model comparison metrics suitable for NOx-limited conditions. Therefore, I recommend the manuscript be accepted for publication. While I think the manuscript could be published without any major changes, I am providing several detailed comments below for the authors to consider for possible additions or clarifications to the analysis in the manuscript. I also suggest that the authors modify Table 4 to include the percent change in daily net O3 production, as described in more detail in the comment below.
Line 407: "The increase in OPE-CPA at low NOx counterintuitively occurs even as O3 production decreases, and the relative changes are notably different, e.g., the OPE-CPA percent increase is 2 times larger than the O3 percent decrease at 50% ANOx. This highlights the difficulty of using OPE to predict O3 response to NOx."
Some clarification of this statement would be useful. Â I don't find this result to be counterintuitive because we expect that O3 production will decrease as NOx emissions are reduced (in NOx-limited regimes). We also expect that the O3 production percent decrease will be smaller than the NOx emissions percent decrease because OPE-CPA will be greater at the higher VOC/NOx ratios associated with the NOx emissions decrease.
Also, given that background O3 is a large contributor to the modeled O3 concentration, we expect that the O3 concentration percent decrease to be smaller than the O3 production percent decrease. (Although the authors constrain the model such that lateral boundary conditions do not contribute to background O3, the residual O3 in layer two and carry over of O3 in layer one effectively acts as a source of high background O3 for each day of the simulation). In diagnostic model evaluations, it is important to distinguish between changes in O3 concentration versus changes in O3 production. If you modeled the relationship of OPE-CPE as a function of VOC/NOx, I expect that OPE-CPA would be more useful in predicting the O3 production decrease for a given NOx emissions decrease, although other feedback processes in the chemical mechanism and in the model will still make OPE-CPA an imperfect predictor of the O3 production response to NOx emissions changes.
Line 298: Â Â O3 response surface plots. I noted that the plots are inverted from traditional O3 isopleth plots in which the NOx scale factor is shown on the y-axis and the x-scale factor is shown on the x-axis. It is interesting that the response surface plots do not show significant reductions in ozone associated with NOx saturation even at the highest modeled NOx concentrations. This is consistent with the O3 time-series plots in Figure 2 that show morning minimum O3 concentrations around 65 ppb. Â The formulation of the model is such that it never shows strong NOx titration effects with O3 concentrations approaching zero that are often seen in urban core areas with large early morning rush hour NOx emissions. Â I don't think this is necessarily a problem for the analysis because O3 and NOz production are primarily controlled by the increasing mass in the PBL as its height rapidly increases in mid-morning. Â However, it is worth noting that the modeled O3 time-series and O3 response plots differ from what we typically see in urban surface layer observed O3 data. It would be interesting to see OPE-CPA results for an actual 3D photochemical grid model simulation for Texas to see if the results are similar to this constrained model analysis. The results might be similar in modeled grid cells that are NOx limited, but might differ in areas that are NOx-saturated or transitional from NOx-saturated to NOx-limited.
Line 275: Â Table 4 compares the effect of a 50% NOx emissions reduction on OPE-CPA and peak O3 concentration (averaged over four days). Â It would also be useful to show the change in daily O3 production which is the metric more relevant to OPE-CPA. Â The daily peak O3 concentration is strongly influenced by the background O3 (carry-over from the previous day as noted above). Â From Figure 2, it appears that the daily morning background ozone averages about 63 ppb for all four mechanisms, and the daily net O3 Production varies from about 24 to 28 ppb. I suggest that the authors also calculate the daily net O3 production for the 50% NOx emissions reduction simulation, and include the percent change in daily net O3 production in Table 4.
Â
Line 384: Â "It is unclear why a linear relationship of O3 and NOz is observed in measurements despite a nonlinear relationship between Pn(O3) and Pn(NOz) (Kleinman et al., 2002)"
I can speculate that the measurements are sampling from air parcels with similar VOC/NOx ratios, with O3 and NOz concentration being influenced primarily by dilution of the plume, so the measurement derived OPE reflects a much more compressed range of VOC/NOx ratios as compared to the scale from zero to a factor of 5 changes in the modeled NOx emissions used to generate OPE_CPA in Figure 5. However, as the authors note, more investigation is needed to understand this relationship. Â